- From: Jim Hendler <hendler@cs.umd.edu>
- Date: Tue, 17 Sep 2002 21:57:21 -0400
- To: pat hayes <phayes@ai.uwf.edu>
- Cc: www-webont-wg@w3.org
At 7:18 PM -0500 9/17/02, pat hayes wrote: >Can I suggest that in this discussion thread we try to avoid using >the word 'definition'? There really is no such thing as a definition >in RDF or OWL. There are only assertions; one can never know that >one has got *all* the relevant information about something. So the >idea of 'containing' some chunk of information to be the >*definition* of a class is beside the point here; its like designing >a trap for unicorns. > >Seems to me that a lot of the discussion about importing has really >been about what counts as an ontology. Let me suggest that we decide >this at the syntactic level by saying that an ontology is a set (or >possibly a bag) of RDF triples. That is not saying that every set of >RDF triples counts as an ontology, but that the triples-store is the >appropriate basic syntactic level for defining such things as >identity of ontologies, mergings of ontologies, entailments between >ontologies and so on. This is really just following the RDF spec >itself. actually I'm fine with this but... >Before the howling starts, let me give some arguments for this. >First, we have decided that the interchange syntax for OWL is >RDF/XML. But RDF/XML is not a suitable notation for defining >RDF-meaningful syntactic operations on: that is, RDF-meaningful >notions of merging, containment (of one set of assertions in >another) and so on do not correspond to simple syntactic operations >on the XML surface syntax. So the aforementioned decision about >RDF/XML only makes sense, seems to me, if we agree that this >interchange language is in fact being used in the way described by >the RDF spec itself, ie as a surface/interchange notation for RDF >*graphs*. A possible objection to this interpretation has always >been that the OWL semantics does not agree with the RDF semantics >when applied to OWL/RDF, so this RDF-centric perspective is not >viable when one wishes to consider semantically meaningful >operations on OWL: I think that objection is now refuted, or at >least has been demoted from a technical objection to an aesthetic >one, so should be discounted. Which brings me to the second point: >treating the RDF graph syntax as the basic syntactic level allows us >to fairly cleanly define OWL-meaningful operations on ontologies, in >a uniform way with how RDFS-meaningful operations are defined on >them. And third, this approach preserves the desired >interoperability and overall coherence between RDF, RDFS and OWL >that we all pray for every evening, right? ok, but... >On this view, then, having [imports B] included in A would be saying >(semantically) that if if the graph-merge of A and B entails C then >A entails C, or (syntactically) that A should be considered to have >B graph-merged into it; where 'A' and 'B' throughout are taken to >refer to whatever collections of RDF triples the syntactic form >being used maps into. In the case of OWL/RDF/XML, that would an OWL >closure, which might be quite a large set: OK, but that's OWL life. >Implemeters of course can choose to be clever in various ways. > I'm even okay with this, however what I have a problem with is the following At URI1: .... <owl:class rdf:ID="foo" /> (1,000,000 other assertions that appear in the graph) At URI2: <:bar owl:subclass URI1:foo /> (put in any owl:ontology and rdf:RDF syntax you want - but no owl:imports in URI2:) In this case I have a real problem with merging the graphs -- the user is very unlikely to actually intend that those million facts which he or she may not even have read should be included. This is the case I really care about. For imports anything that can identify and merge graphs makes me happy - for this case, I care that we somehow scope what is included. I would like this to have the same semantics as having one URI which contained <owl:class rdf:ID="foo" /> <:bar owl:subclass :foo /> (i.e. nothing else from URI1: is to be included unless it is explicitly mentioned.) -- Professor James Hendler hendler@cs.umd.edu Director, Semantic Web and Agent Technologies 301-405-2696 Maryland Information and Network Dynamics Lab. 301-405-6707 (Fax) Univ of Maryland, College Park, MD 20742 240-731-3822 (Cell) http://www.cs.umd.edu/users/hendler
Received on Tuesday, 17 September 2002 21:57:25 UTC